Related Experiment Videos
Diagrammatic dataset on AI-generated formative feedback for XML-based UML models
Janka Pecuchová1, Ľubomír Benko1, Martin Drlík1
1Faculty of Natural Sciences and Informatics, Constantine the Philosopher University in Nitra, 949 01 Nitra, Slovakia.
Data in Brief
|June 30, 2026
Summary
This dataset offers student UML models and AI feedback in Slovak, enabling research on automated assessment and large language model performance in non-English software engineering education.
Area of Science:
- Computer Science Education
- Artificial Intelligence in Education
- Software Engineering Pedagogy
Background:
- Educational datasets are crucial for advancing AI in learning.
- Software Engineering education benefits from automated feedback systems.
- Multilingual applications of AI in education require specific datasets.
Purpose of the Study:
- To release a novel dataset of student-generated UML models and AI-generated formative feedback.
- To facilitate research on automated formative assessment in software engineering.
- To enable studies on the effectiveness of large language models in non-English educational contexts.
Main Methods:
- Collected and anonymized 112 student records, 448 feedback records, and 700 XML reports.
- Utilized Enterprise Architect (v16) for XML model generation.
- Employed OpenAI's GPT-4-Turbo API for generating Slovak-language formative feedback.
Main Results:
- The dataset links UML models, AI feedback, scores, grades, and human evaluations.
- Provides insights into AI feedback generation in Slovak for technical disciplines.
- Offers a reproducible basis for benchmarking and extending research in software engineering education.
Conclusions:
- The dataset supports research on automated formative assessment and prompt engineering.
- Facilitates human-AI feedback comparison and multilingual feedback analysis.
- Highlights the potential of LLMs in diverse educational settings and languages.
